Pith. sign in

REVIEW 5 major objections 5 minor 30 references

The role of gain neuromodulation in layer-5 pyramidal neurons

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Bursting in layer-5 pyramidal neurons accelerates spike-timing-dependent plasticity in near-linear proportion to gain, suggesting neuromodulatory gain pulses are an adaptive two-timescale learning mechanism.

desk verdict A plausible computational demonstration that gain modulation speeds STDP update rates, but the headline result is largely a built-in consequence of the eligibility-trace rule, not a tested burst-specific effect. read the letter →

arxiv 2507.03222 v2 pith:LQUAGR3R submitted 2025-07-03 q-bio.NC cs.AI

Alejandro Rodriguez-Garcia , Christopher J. Whyte , Brandon R. Munn , Jie Mei , James M. Shine , Srikanth Ramaswamy This is my paper · ORCID
classification q-bio.NCcs.AI
keywords layer-5pyramidalneuronsgainneuromodulationcalciumplateaupotentialssoma-apicalcouplingburstfiringspike-timing-dependentplasticityfast-slowweightdecompositionneuromodulatorypulses
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that neuromodulatory control of layer-5 pyramidal neuron gain is a two-timescale learning mechanism. In a spiking network model with two-compartment pyramidal neurons, stronger soma–apical coupling or stronger apical dendritic drive makes neurons burst more often, and bursting raises the slope of the input–output curve. The study then shows that the same manipulation accelerates the rate of spike-timing-dependent plasticity (STDP) weight updates in an essentially linear way ($r=0.97$), and that a brief neuromodulator-like pulse of gain can transiently speed synaptic change without invoking an explicit reward signal. The authors propose that such gain pulses let a circuit toggle between fast, flexible reconfiguration and slow, stable retention, helping to resolve the plasticity–stability dilemma.

What carries the argument

The argument is carried by a two-compartment spiking-neuron model: an Izhikevich somatic compartment coupled probabilistically to an apical compartment whose regenerative nonlinearity creates bistable calcium plateau potentials. The soma–apical coupling parameter sets the probability that each somatic spike backpropagates to the dendrite; when apical drive is strong enough, the plateau changes the somatic reset and adaptation variables so that a single somatic spike becomes a burst. A Gaussian-connectivity network of these pyramidal cells with somatostatin (SOM) and parvalbumin (PV) interneurons runs the eligibility-trace STDP rule, and the appendix adds a temporal low-pass filter that splits each synaptic weight into slow and fast components.

What would settle it

If, in a layer-5 pyramidal cell recorded in vitro, the weight-change rate per spike pair is identical for regular spiking and burst firing when total spike count is matched (for example using dynamic clamp to substitute plateau-triggered bursts), then the near-linear acceleration of STDP by coupling and apical drive would be an artifact of the firing-rate dependence of the assumed plasticity rule rather than a distinct burst effect.

Watch

Extended reading notes

Core claim

The central claim is that gain modulation in layer-5 pyramidal neurons is carried by calcium-plateau-driven bursting. Somatic backpropagating action potentials and distal dendritic drive interact through a bistable apical compartment; when apical voltage crosses roughly $-30$ mV, the somatic reset and adaptation parameters switch the neuron from regular spiking to bursting. Bursting increases the gain of the somatic input–output function, and because the eligibility-trace STDP rule counts spike pairs, the increase in spiking translates into a near-linear increase in the absolute rate of synaptic weight updates ($r=0.97$ for both coupling and apical-drive sweeps). Dendritic-targeting inhibition suppresses gain by preventing plateau generation, while somatic-targeting inhibition raises the firing threshold and gates output. The paper concludes that brief phasic neuromodulation of apical drive and soma–apical coupling acts as an adaptive two-timescale optimizer, effectively rescaling the learning rate at the synapse.

Load-bearing premise

The load-bearing premise is that the eligibility-trace STDP rule, in which every spike pair contributes an update, reflects how real synapses change, so that a burst-driven rise in firing rate is taken to mean that real synapses update faster; the paper does not test this mapping, nor does it validate the assumed link from neuromodulators to the two parameters it sweeps.

Editorial extensions

If this is right

  • If the model is right, a brief neuromodulatory pulse that raises apical drive or soma–apical coupling transiently increases gain and STDP update rates, allowing rapid synaptic reconfiguration during states of uncertainty or arousal.
  • The same mechanism can dial the network back to low gain and slow weight changes, providing a stability mode that preserves existing connectivity.
  • Dendritic inhibition and somatic inhibition play distinct roles: dendritic inhibition removes gain without strongly suppressing regular spiking, while somatic inhibition gates whether the cell fires at all.
  • Because bursting acts as a coincidence detector for basal and apical input, the model links gain modulation to predictive-coding-like integration of top-down and bottom-up signals.
  • The fast–slow weight decomposition derived in the appendix suggests that phasic gain changes create memories that update quickly without erasing slow, stable knowledge, relevant to continual learning.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct experimental test suggested by the model is whether STDP update rate scales with burst probability rather than with total spike count; a dynamic-clamp preparation that keeps spike count constant while varying burst probability could separate the two.
  • Because the near-linear correlation is built into the spike-pair-triggered eligibility-trace rule, part of the claimed acceleration follows from any manipulation that raises firing rate; the biologically novel step is the assumed mapping from acetylcholine and noradrenaline to coupling and apical drive, which the paper leaves unvalidated.
  • If the fast–slow decomposition generalizes, brief gain pulses could act as an implicit learning-rate schedule, suggesting that neuromorphic hardware could emulate arousal-like gain modulation to alternate exploration and consolidation phases.
  • The model predicts that optogenetic activation of SOM interneurons should slow STDP updates by suppressing bursts, while PV activation should gate updates by raising threshold; paired whole-cell recordings could test this prediction.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper presents a two-compartment Izhikevich model of layer-5 pyramidal neurons, SOM interneurons, and PV interneurons, and uses it to study how soma-apical coupling and apical dendritic drive modulate neuronal gain. The authors show qualitatively that stronger coupling or drive increases gain and burst ratio, that dendritic inhibition suppresses bursting while somatic inhibition raises firing threshold, and that increasing gain raises the rate of STDP weight updates. They argue that brief neuromodulatory gain pulses can therefore act as a two-timescale optimization mechanism, with a fast component incorporating new evidence and a slow component preserving existing knowledge. The central numerical claims are the near-linear correlations of r=0.97 between coupling or apical drive and STDP update rate (Figure 5a,c) and the fast-slow decomposition in Appendix A. The manuscript is a short conference-style paper and does not provide code or data.

Significance. If the qualitative results hold, the model is a computationally efficient substrate for BAC firing and provides a concrete link between ascending-arousal neuromodulation, dendritic gain, and plasticity. The reproduction of known effects, such as compartment-specific inhibition and gain enhancement by apical drive, is a useful sanity check. However, the novelty of the central learning claim is limited: the STDP update rate is, by construction of Eqs. (2)-(5), proportional to the number of pre/post spike pairs, so manipulations that increase firing rate will necessarily increase update rate. The fast-slow decomposition in Appendix A is an algebraic identity rather than an emergent two-timescale mechanism. The paper also ships no code or data, which constrains verification of the quantitative figures. These issues do not invalidate the qualitative gain-modulation story, but they do require the authors to reframe or test the central claim more carefully.

major comments (5)
  1. [§3.4, Eqs. (2)–(5)] The claim that bursting accelerates STDP is not dissociated from firing-rate increases. In the plasticity rule, P± is incremented at every pre/postsynaptic spike, and z is updated at every post/presynaptic spike, so the total eligibility signal is proportional to the integral of the spike-pair count. Figures 3 and 5 show that higher coupling or apical drive raises firing rate and burst ratio; hence the near-linear r=0.97 relations in Figure 5a,c are expected from the rate increase alone. The authors should add a rate-matched control: with mean firing rate held constant, does a higher burst fraction still increase the STDP update rate? Without such a control, the manuscript over-interprets 'bursting accelerates STDP' and the resulting optimization story.
  2. [§3.1, Eq. (11)] The nullcline derivation in Eq. (11) is malformed: it reads as if the v-nullcline and w-nullcline are being set equal to each other and also to a third expression, which is not a valid derivation. The authors should write the two nullclines separately, v̇_d = 0 and ẇ_d = 0, and then describe the intersection points. This matters because the bistability analysis in Figure 2 depends on the nullcline structure, but the current equation does not permit the reader to check the claimed saddle-node bifurcations.
  3. [Appendix A, Eqs. (13)–(21)] The fast-slow decomposition is presented as a substantive result, but it is a formal identity. For any signal Γ, defining Γ0 as its low-pass filter and then defining w_slow and w_fast with the rates in Eqs. (20)-(21) guarantees that w = w_slow + w_fast. The appendix does not show that gain pulses actually produce two timescales in the simulated synapses, nor does it test whether the two components evolve with distinct, robust timescales. The authors should either simulate the decomposition and report the actual timescales of w_slow and w_fast under a gain pulse, or explicitly state that this is bookkeeping rather than a mechanistic finding.
  4. [Figure 5 and §3.4] The quantitative support for the central claim is thin: the reported r=0.97 correlations are given without confidence intervals, and despite the claim that 100 trials were run, no error bars are shown. The caption states that error bars are omitted because 'the dot size is bigger than their magnitude', but this is not verifiable from the figure. The authors should provide standard error bars or confidence intervals, or explicitly state that the figure shows trial means and report the variance across trials.
  5. [Reproducibility] The manuscript does not state where the simulation code, parameter files, or data are available. Since all figures are generated from a custom model with many parameters, and since the central quantitative claims depend on the exact implementation of the OU drive, BAP probability, and STDP update rule, the absence of code or data makes the numerical results unverifiable. I strongly recommend depositing the code and all parameter settings in a public repository.
minor comments (5)
  1. [Throughout] There are numerous typographical errors, including 'Celular', 'numercial', 'sompatic', 'prompmting', 'betuween', and 'paraments'; these should be corrected in a revised version.
  2. [Figure 5 caption] The caption states '2000 pA of basal drive', while the text of §3.4 mentions a mean basal drive of 450 pA for panel a; please clarify which value applies to which panel and ensure the caption and text are consistent.
  3. [§3.1, Eq. (12)] The definition of I_e^d in Eq. (12) as m_d P_BAP H + I_d does not include the soma-apical coupling parameter that is swept later in the paper; it would be helpful to state explicitly how this expression relates to the coupling strength used in Figures 3 and 5.
  4. [Figure 2] The bifurcation diagram reports specific current values (538.91 pA and 647.37 pA) but the derivation of these values from the nullclines is not shown; please provide the parameter values or the calculation.
  5. [§3.4, Figure 5d] The caption refers to 'top panel', but the figure has three subpanels in (d); please label the subpanels explicitly and reference them by their labels in the text.

Circularity Check

2 steps flagged · score 7.0 of 10

The two central plasticity conclusions—'bursting accelerates STDP' and the fast–slow weight decomposition—are by-construction consequences of the eligibility-trace rule and of the low-pass filter identity.

  1. self definitional [Methods Sec. 2.1, Eqs. (4)–(5); Results Sec. 3.4, Fig. 5a,c]
    "dzi j/dt =− zi j/τz + 1/τz [ P+ i j Σ_m δ(t−t_j^(m)) + P− i j Σ_k δ(t−t_i^(k)) ] ,(4) dwi j/dt =ηz i j. (5) ... We hypothesize that if synaptic plasticity depends on the spiking behavior of neurons, then synaptic updates occur more rapidly at higher firing rates."

    Equation (4) increments the eligibility variable z at every pre-/post-synaptic spike pair (terms P+Σδ(t−tj) + P−Σδ(t−ti)), and Eq. (5) sets dw/dt=ηz. Hence the integrated number of STDP updates is, to first order, proportional to the number of spike pairs. Figures 3 and 5 show that raising soma–apical coupling or apical drive increases firing rate and burst ratio, so the near-linear r=0.97 update-rate increases in Fig. 5a,c are the rule's built-in rate dependence, not a burst-specific learning effect. The paper provides no rate-matched control (fixed mean firing rate, varying burst fraction) to dissociate bursting from firing rate. The 'prediction' is therefore a definitional consequence of the plasticity rule, not an independent model discovery.

  2. self definitional [Appendix A, Eqs. (13)–(21); Outlook Sec. 4]
    "We then decompose the update rate as Γ(w(t),t)=Γ0(t)+[Γ(w(t),t)−Γ0(t)], (17) ... Finally, defining w(t)=wslow(t)+wfast(t), (19) and imposing d wslow/dt=ηΓ0(t), (20) d wfast/dt=η[Γ(w,t)−Γ0(t)], (21) we obtain two coupled dynamics whose sum reproduces the full update."

    This is an algebraic identity for any plasticity signal Γ: w_slow is the integral of a low-pass-filtered version of Γ, and w_fast is the residual; w_slow+w_fast equals w by construction (Eqs. 17–21). No property of the STDP model, burst dynamics, or gain modulation enters. The Outlook's claim that 'phasic plasticity changes causes each synaptic weight to decompose into fast and slow components' and the proposed two-timescale optimization mechanism therefore add nothing beyond renaming a low-pass filter decomposition; any rule dw/dt=Γ admits the same split. This is self-definitional.

full rationale

The paper contains substantial non-circular modeling: the calcium-plateau bistability analysis (Eq. 11), the input-output gain curves, and the compartment-specific inhibition effects are genuine simulations with stated equations and parameters. The use of prior work by the same authors (Whyte et al. 2025, Munn et al. 2023) to justify the dual-compartment architecture is self-citation, but it is not the load-bearing step for the plasticity claims. The circularity is concentrated in two central plasticity conclusions. First, 'bursting accelerates STDP' (Abstract, Section 3.4) is not an independent result: with the eligibility-trace rule of Eqs. (2)–(5), z is incremented at every spike pair and dw/dt=ηz, so any manipulation that raises firing rate (as coupling and apical drive do in Fig. 3) must raise the update count. The reported r=0.97 correlations (Fig. 5a,c) therefore quantify the rule's built-in rate dependence; no rate-matched burst-vs-tonic control is performed. Second, the 'fast-slow weight decomposition' (Appendix A, Eqs. 17–21) is an algebraic identity: w_slow integrates a low-pass-filtered plasticity rate and w_fast is the residual, so w_slow+w_fast reproduces w for any plasticity rule by construction. The conclusion that phasic gain pulses 'decouple every synaptic weight into a fast-slow regime' and thus enable continual learning is a restatement of the filter decomposition, not a derivation from the simulated network. These by-construction steps make the paper's headline plasticity/optimization claims partially circular, despite the independent gain-modulation results.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The model depends on many hand-chosen or inherited parameters. The specific values used for STDP, connectivity, and dendritic dynamics are not fully disclosed, making it hard to assess how much of the result is built into the choices. No new physical entities are postulated; the fast-slow weight decomposition is a mathematical rearrangement of existing variables.

free parameters (5)
  • Soma-apical coupling strength = swept 0 to 1 in Fig 3a
    Central variable for gain modulation; not fit to data, chosen as model parameter.
  • Mean apical dendritic drive (µ_i for dendrite) = swept 200 to 600 pA
    Controls Ca2+ plateau generation; chosen to illustrate gain changes.
  • STDP eligibility trace time constant τ_z = not given
    STDP update rate result depends on this; value taken from prior work but not specified in the paper.
  • STDP learning rate η = not given
    Scales weight updates; no value provided in text.
  • Connectivity parameters C_E, C_I, d_E, d_I = not given
    Shape synaptic weights; values not reported.
assumptions (5)
  • domain assumption The Izhikevich quadratic adaptive integrate-and-fire model adequately represents the somatic dynamics of layer-5 pyramidal neurons, PV interneurons, and SOM interneurons.
    Used in Section 2.2 (Eq. 7). The model is standard, but the two-compartment extension is an unvalidated simplification.
  • domain assumption STDP with eligibility traces (Eqs. 2-5) is a valid model of synaptic plasticity in this circuit.
    The paper's main claim about accelerated learning is a direct consequence of this rule.
  • ad hoc to paper The apical dendritic compartment exhibits a sigmoidal regenerative nonlinearity (Eq. 10) and bistable Ca2+ plateaus for the chosen parameters.
    The nonlinearity is a modeling choice from prior work; the bistability is demonstrated via a bifurcation diagram but the printed derivation in Eq. (11) is malformed.
  • domain assumption Synaptic connectivity follows a Gaussian profile with periodic boundary conditions (Eq. 1).
    Simplifies network wiring; values for C_E, C_I, d_E, d_I are not reported.
  • domain assumption Backpropagating action potentials occur probabilistically with probability P_BAP.
    Introduced in Section 2.2, based on experimental observations [18], but the probability is not quantified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The role of gain neuromodulation in layer-5 pyramidal neurons." pith.science (2026). https://pith.science/paper/LQUAGR3R

@misc{pith2026250703222,
  author       = {Pith},
  title        = {Pith review of: The role of gain neuromodulation in layer-5 pyramidal neurons},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LQUAGR3R}},
  note         = {Machine review of arXiv:2507.03222}
}
read the original abstract

Biological and artificial learning systems alike confront the plasticity-stability dilemma. In the brain, neuromodulators such as acetylcholine and noradrenaline relieve this tension by tuning neuronal gain and inhibitory gating, balancing segregation and integration of circuits. Fed by dense cholinergic and noradrenergic projections from the ascending arousal system, layer-5 pyramidal neurons in the cerebral cortex offer a relevant substrate for understanding these dynamics. When distal dendritic signals coincide with back-propagating action potentials, calcium plateaus turn a single somatic spike into a high-gain burst, and interneuron inhibition sculpts the output. These properties make layer-5 cells gain-tunable amplifiers that translate neuromodulatory cues into flexible cortical activity. To capture this mechanism we developed a two-compartment Izhikevich model for pyramidal neurons and single-compartment somatostatin (SOM) and parvalbumin (PV) interneurons, linked by Gaussian connectivity and spike-timing-dependent plasticity (STDP). The soma and apical dendrite are so coupled that somatic spikes back-propagate, while dendritic plateaus can switch the soma from regular firing to bursting by shifting reset and adaptation variables. We show that stronger dendritic drive or tighter coupling raise gain by increasing the likelihood of calcium-triggered somatic bursts. In contrast, dendritic-targeted inhibition suppresses gain, while somatic-targeted inhibition raises the firing threshold of neighboring neurons, thus gating neurons output. Notably, bursting accelerates STDP, supporting rapid synaptic reconfiguration and flexibility. This suggests that brief gain pulses driven by neuromodulators could serve as an adaptive two-timescale optimization mechanism, effectively modulating the synaptic weight updates.

Figures

Figures reproduced from arXiv: 2507.03222 by the authors.

Figure 1
Figure 1. Study of layer-5 neurons with PV and SOM inhibition. a) Schematic of the model, illustrating apical activity (Ca2+ spikes) and somatic activity (bursts and isolated spikes) in isolation. b) Single neuron dynamics of each neuronal type considered in the network: apical and sompatic compartments of layer 5 pyramidal neurons in light red, fast spiking PV interneurons in dark blue and regular spiking SOM interneurons in… view at source ↗
Figure 2
Figure 2. Calcium plateaus generation via apical dendritic bistability. (a) Bifurcation diagram of the layer-5 pyramidal neuron apical compartment as the apical drive I e d is varied. A saddle-node bifurcation at Ie(1) d = 538.91pA gives rise to a stable plateau potential coexisting with the resting state. At a second saddle-node bifurcation Ie(2) d = 647.37pA the resting state disappears and the plateau potential becomes the… view at source ↗
Figure 3
Figure 3. Modulation of somatic responses by apical-somatic coupling and apical dendritic drive in layer 5 pyramidal neurons. Effect of apical-somatic coupling: Simulations were performed with a OU apical drive of 400 pA, sufficient to shift the stability point of the dendritic nonlinearity via backpropagating action potentials (BAPs). a) Somatic firing rate curves for varying apical-somatic coupling strengths show that incre… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Suppression of somatic responses by inhibitory control at proximal and apical dendritic compartments in layer-5 pyramidal neurons. Effect of proximal (somatic) inhibition: Simulations were performed with a OU apical drive of 500 pA and fully coupled neurons, to ensure …
Figure 5
Figure 5. Figure 5: Gain neuromodulation enhances STDP update rates: Simulations were made with a 100 ms delay, 2000 pA of basal drive, and a drive to SOM inhibitory neurons of 600 pA and none to PV interneurons. a) In this case, we used 450 pA of apical drive and sweeped the coupling par…
Figure 6
Figure 6. Figure 6: Modulation of somatic responses by apical-somatic coupling and apical dendritic drive in a network of layer 5 pyramidal neurons. Effect of apical-somatic coupling: Simulations were performed with a OU apical drive of 450 pA, 600 pA to SOM neurons and 0 pA to PV interne…
Figure 7
Figure 7. Figure 7: Inhibition of layer 5 pyramidal neurons. a) Effect of proximal (somatic) inhibition on the input-output curve of pyramidal neurons (’shunting effect’). b) Effect of apical (dendritic) inhibition on the input-output curve of pyramidal neurons (’gain modulation’). Simula…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 25 canonical work pages

  1. [1]

    E., Zhu, J

    Larkum, M. E., Zhu, J. J. & Sakmann, B. A new cellular mechanism for coupling inputs arriving at different cortical layers.Nature398, 338–341, DOI: 10.1038/18686 (1999)

    work page doi:10.1038/18686 1999

  2. [2]

    Larkum, M. E. Top-down Dendritic Input Increases the Gain of Layer 5 Pyramidal Neurons.Cereb. Cortex14, 1059–1070, DOI: 10.1093/cercor/bhh065 (2004)

    work page doi:10.1093/cercor/bhh065 2004

  3. [3]

    & Sejnowski, T

    Salinas, E. & Sejnowski, T. J. Correlated neuronal activity and the flow of neural information.Nat. Rev. Neurosci.2, 539–550, DOI: 10.1038/35086012 (2001)

    work page doi:10.1038/35086012 2001

  4. [4]

    Wainstein, G.et al.Gain neuromodulation mediates task-relevant perceptual switches: evidence from pupillom- etry, fMRI, and RNN Modelling, DOI: 10.7554/eLife.93191.2 (2025)

    work page doi:10.7554/elife.93191.2 2025

  5. [5]

    J.et al.A burst-dependent thalamocortical substrate for perceptual awareness.PLOS Comput

    Whyte, C. J.et al.A burst-dependent thalamocortical substrate for perceptual awareness.PLOS Comput. Biol. 21, e1012951, DOI: 10.1371/journal.pcbi.1012951 (2025)

    work page doi:10.1371/journal.pcbi.1012951 2025

  6. [6]

    R.et al.Neuronal connected burst cascades bridge macroscale adaptive signatures across arousal states.Nat

    Munn, B. R.et al.Neuronal connected burst cascades bridge macroscale adaptive signatures across arousal states.Nat. Commun.14, 6846, DOI: 10.1038/s41467-023-42465-2 (2023)

    work page doi:10.1038/s41467-023-42465-2 2023

  7. [7]

    Shine, J. M. Neuromodulatory control of complex adaptive dynamics in the brain.Interface F ocus.13, 20220079, DOI: 10.1098/rsfs.2022.0079 (2023)

    arXiv 2023

  8. [8]

    & Szabadi, E

    Samuels, E. & Szabadi, E. Functional Neuroanatomy of the Noradrenergic Locus Coeruleus: Its Roles in the Regulation of Arousal and Autonomic Function Part II: Physiological and Pharmacological Manipulations and Pathological Alterations of Locus Coeruleus Activity in Humans.Curr. Neuropharmacol.6, 254–285, DOI: 10.2174/157015908785777193 (2008)

    work page doi:10.2174/157015908785777193 2008

Show all 30 references
  1. [9]

    Labarrera, C.et al.Adrenergic Modulation Regulates the Dendritic Excitability of Layer 5 Pyramidal Neurons In Vivo.Cell Reports23, 1034–1044, DOI: 10.1016/j.celrep.2018.03.103 (2018)

    2018 doi

  2. [10]

    Williams, S. R. & Fletcher, L. N. A Dendritic Substrate for the Cholinergic Control of Neocortical Output Neurons.Neuron101, 486–499.e4, DOI: 10.1016/j.neuron.2018.11.035 (2019)

    2019 doi

  3. [11]

    R.et al.A thalamocortical substrate for integrated information via critical synchronous bursting

    Munn, B. R.et al.A thalamocortical substrate for integrated information via critical synchronous bursting. Proc. Natl. Acad. Sci.120, e2308670120, DOI: 10.1073/pnas.2308670120 (2023)

    2023 doi

  4. [12]

    Florian, R. V . Reinforcement Learning Through Modulation of Spike-Timing-Dependent Synaptic Plasticity. Neural Comput.19, 1468–1502, DOI: 10.1162/neco.2007.19.6.1468 (2007)

    2007 doi

  5. [13]

    Simple model of spiking neurons.IEEE Transactions on Neural Networks14, 1569–1572, DOI: 10.1109/TNN.2003.820440 (2003)

    Izhikevich, E. Simple model of spiking neurons.IEEE Transactions on Neural Networks14, 1569–1572, DOI: 10.1109/TNN.2003.820440 (2003)

    2003

  6. [14]

    & Ramaswamy, S

    Rodriguez-Garcia, A., Mei, J. & Ramaswamy, S. Enhancing learning in spiking neural networks through neuronal heterogeneity and neuromodulatory signaling, DOI: 10.48550/ARXIV .2407.04525 (2024). Version Number: 4

    2024 doi

  7. [15]

    Gast, R., Solla, S. A. & Kennedy, A. Neural heterogeneity controls computations in spiking neural networks. Proc. Natl. Acad. Sci.121, e2311885121, DOI: 10.1073/pnas.2311885121 (2024)

    2024 doi

  8. [16]

    M.Dynamical Systems in Neuroscience: The Geometry of Excitability and Bursting(The MIT Press, 2006)

    Izhikevich, E. M.Dynamical Systems in Neuroscience: The Geometry of Excitability and Bursting(The MIT Press, 2006)

    2006

  9. [17]

    & Sprekeler, H

    Naud, R. & Sprekeler, H. Sparse bursts optimize information transmission in a multiplexed neural code.Proc. Natl. Acad. Sci.115, DOI: 10.1073/pnas.1720995115 (2018)

    2018 doi

  10. [18]

    M.et al.The stochastic nature of action potential backpropagation in apical tuft dendrites.J

    Short, S. M.et al.The stochastic nature of action potential backpropagation in apical tuft dendrites.J. Neurophysiol.118, 1394–1414, DOI: 10.1152/jn.00800.2016 (2017)

    2017

  11. [19]

    Version Number: 1

    Mei, J.et al.Improving the adaptive and continuous learning capabilities of artificial neural networks: Lessons from multi-neuromodulatory dynamics, DOI: 10.48550/ARXIV .2501.06762 (2025). Version Number: 1

    2025 doi

  12. [20]

    & Larkum, M

    Suzuki, M. & Larkum, M. E. General Anesthesia Decouples Cortical Pyramidal Neurons.Cell180, 666– 676.e13, DOI: 10.1016/j.cell.2020.01.024 (2020)

    2020 doi

  13. [21]

    Viewpoint: the core and matrix of thalamic organization.Neuroscience85, 331–345, DOI: 10.1016/ S0306-4522(97)00581-2 (1998)

    Jones, E. Viewpoint: the core and matrix of thalamic organization.Neuroscience85, 331–345, DOI: 10.1016/ S0306-4522(97)00581-2 (1998)

    1998

  14. [22]

    Sherman, S. M. Thalamocortical interactions.Curr. Opin. Neurobiol.22, 575–579, DOI: 10.1016/j.conb.2012. 03.005 (2012)

    2012 doi

  15. [23]

    & Arsenault, D

    Zhang, Z.-w. & Arsenault, D. Gain modulation by serotonin in pyramidal neurones of the rat prefrontal cortex. The J. Physiol.566, 379–394, DOI: 10.1113/jphysiol.2005.086066 (2005)

    2005 arXiv

  16. [24]

    P., Bengio, Y

    Sacramento, J., Costa, R. P., Bengio, Y . & Senn, W. Dendritic cortical microcircuits approximate the backprop- agation algorithm, DOI: 10.48550/ARXIV .1810.11393 (2018). Version Number: 1

    2018 doi arXiv

  17. [25]

    W., Pemberton, J., Mellor, J

    Greedy, W., Zhu, H. W., Pemberton, J., Mellor, J. & Costa, R. P. Single-phase deep learning in cortico-cortical networks, DOI: 10.48550/ARXIV .2206.11769 (2022). Version Number: 2

    2022 doi arXiv

  18. [26]

    Version Number: 1

    Aizenbud, I.et al.Neural mechanisms of predictive processing: a collaborative community experiment through the OpenScope program, DOI: 10.48550/ARXIV .2504.09614 (2025). Version Number: 1

    2025 doi

  19. [27]

    Strack, B., Jacobs, K. M. & Cios, K. J. Biological restraint on the Izhikevich neuron model essential for seizure modeling. In2013 6th International IEEE/EMBS Conference on Neural Engineering (NER), 395–398, DOI: Accepted at 34th Annual Computational Neuroscience Meeting Rodri...

  20. [28]

    & Ramaswamy, S

    Colangelo, C., Shichkova, P., Keller, D., Markram, H. & Ramaswamy, S. Cellular, Synaptic and Network Effects of Acetylcholine in the Neocortex.Front. Neural Circuits13, 24, DOI: 10.3389/fncir.2019.00024 (2019)

    2019

  21. [29]

    & Trevelyan, A

    Pouille, F., Watkinson, O., Scanziani, M. & Trevelyan, A. J. The contribution of synaptic location to inhibitory gain control in pyramidal cells.Physiol. Reports1, DOI: 10.1002/phy2.67 (2013)

    2013 doi

  22. [30]

    slow” (tonic) and “fast

    Hinton, G. E. & Plaut, D. C. Using fast weights to deblur old memories. InProceedings of the Ninth Annual Conference of the Cognitive Science Society, 177–186 (Hillsdale, NJ: Erlbaum, 1987). Accepted at 34th Annual Computational Neuroscience Meeting Rodriguez-Garcia et al. (Pr...

    1987

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.